Study of water content control systems for short-lived vegetables using neural networks

Abstract

          The purpose of this research was to improve the efficacy of water use in agriculture especially in the areas with limited water sources. This research focused on how to decrease amount of water to the Chinese kale, one of the most popular vegetable plant of Suphanburi, while the crop production was unaffected. The experiments were done in May and June at Samchuk district, Suphanburi province. The variety of Maejo 1 of Chinese kale was used in this research. The seeds were sowed into two beds of size 2×3 meters located in the mosquito net curved roof. The first bed treated the water by using a timer, while the other bed treated the water by using a neural network. For the first bed, after the calculation using Penman-Monteith formula, 26.08 and 21.60 liters of water per day are used in May and June accordingly. However, in the other bed that used a neural network water treatment, the Penman-Monteith formula and data from the sensors every 7 days, are used to forecast the required amount of water. After the calculation, the second bed required less amount of water than the first bed by 53.78 liters per week in May and 51.54 liters per week in June. However, the average weight of the kales produced from two methods were slightly different by 0.02 gram only. Therefore, the neural network could be used to reduce the amount of treatment water.